<scp>ParsiNLU</scp>: A Suite of Language Understanding Challenges for Persian

Abstract Despite the progress made in recent years addressing natural language understanding (NLU) challenges, majority of this remains to be concentrated on resource-rich languages like English. This work focuses Persian language, one widely spoken world, and yet there are few NLU datasets available for language. The availability high-quality evaluation is a necessity reliable assessment different tasks domains. We introduce ParsiNLU, first benchmark that includes range tasks—reading comprehension, textual entailment, so on. These collected multitude ways, often involving manual annotations by native speakers. results over 14.5k new instances across 6 distinct tasks. Additionally, we present state-of-the-art monolingual multilingual pre-trained models compare them with human performance, which provides valuable insights into our ability tackle challenges Persian. hope ParsiNLU fosters further research advances understanding.1